Mastering

Mastering requires the following ingredients: a good room, good speakers and good processing. But even more, mastering is about making all the right choices when developing the potential of a particular tune: Choose static and dynamic spectral balance. Choose average/peak ratio. Choose spatial tuning. Choose what format to optimize for. Choose what tools to use. Choose to trust your ears instead of trying to top the loudness of another recently released song. Choose music.

Digital Level
In digital production, level has traditionally been measured on a sample-by-sample basis. The highest possible level in a digital encoding is called 0 dBFS (or Full Scale Digital, FSD), and the only thing to be concerned about is not to hit that ceiling with too many samples in a row. However, a signal which needs more headroom to reproduce than a sine wave peaking at 0 dBFS can easily exist in the digital domain. We refer to such signals as "0 dBFS+", and haven’t found a single professional or consumer CD player that doesn’t significantly distort when subjected to them, and many of them even prolong the effect: They latch-up, and take a little while to get out of the distortion mode again, hence a period of time after a peak has occurred will also be distorted.

When CD was invented, analog tapes were typically used for mixing. During mastering, the signal was passed through analog processing such as Compressors and Eqs, and eventually converted to digital. The level and the AD converter headroom were read in the digital domain. Back then, the consecutive sample count method for detecting overloads made sense, because all signals fulfilled the sample theorem fresh out of the AD converter.

Nowadays, the circumstances are different, but the way level is measured remains the same. Audio is digital when it arrives at the mastering studio, and there is no guarantee that the signal doesn’t already contain out of band components from clipping or misbehaved upstream digital processors or workstations. Maybe the mastering engineer even applies an additional arsenal of digital weaponry. If so, she does it blindly, because there are no meters to indicate how polluted the end result will be.

Weapons of Mass Destruction
The lack of intelligent level restriction has turned mastering into a loudness war rather than a quest for getting decent audio quality out of a potentially well sounding media. The war has become so furious that equipment downstream of mastering is not able to pass or reproduce the audio material without adding significant distortion to it, thereby causing early listening fatigue at the end listener.

Digital processors may have familiar names such as Compressor or Limiter, but there is no guarantee that they won’t violate the ground rules of digital audio, and invisibly pollute the signal in ways their analog predecessors would never do.

Disarming our studios
By tracking level history backwards on a number of recent CD releases, we learned that no single piece of equipment in the recording, mixing or mastering chain carries the full responsibility for the current situation.
Digital clipping can happen several places during production: Inside an audio workstation (typically on the mix-buss or in the plug-ins), deliberately in various types of dynamics processors, or even inherited from digital sample libraries. Clipped kick-drums or snares may be chosen for creative reasons, but if the sample is brought close to full scale during mix or mastering, unpredictable reproduction is inevitable.

For a mixing engineer, advice is simple: If you mix to digital, don’t peak higher than -3 dBFS, and let mastering take care of the rest. Alternatively, mix to analog.

For a mastering engineer, the situation is more complicated, but he can routinely check incoming mixes on an oversampled meter. If frequent 0 dBFS+ peaks are present, he should make sure to get rid of them before doing anything else, including sample rate conversion. When setting the level of the final master, keep an eye on more cultivated measures than a sample counter. Oversampled peak meters and histograms are valid visual tools. A limiter such as BrickWall 2, found in Mastering 6000, has the latest research on how to avoid downstream distortion built into it, and it also gives a clear indication of the 0 dBFS+ status.

Loudness
Together with the academic community, TC has produced a new, meaningful loudness measure that can be used on both music and speech, and which can be used for display as well as processing control purposes.

Loudness, unlike level, is subjective and therefore more difficult to measure, especially for time-varying signals like music. Loudness assessment exhibits a certain Between Listener Variability (BLV), even within homogenous groups of listeners, while differences in age, sex, culture etc. can add further to the variation. Individual loudness assessments by the same person are also only consistent to some extent, and it depends on tiredness, mood, the grade of attention etc. This type of variation is called Within Listener Variability (WLV).

A measure of loudness can therefore only be considered of universal value if it is based on large subjective reference tests and reproducible results, processed by appropriate statistical analysis.

TC algorithms are now being based on perceptual measurements rather than a simple peak or time-averaged dose estimate such as variations over the Leq theme.

Integrated Processing, Metering and Monitoring
To better preserve nuances in the audio, and to bring out low-level detail rather than squashing the peaks, you should consider starting to use new types of dynamics processing such as DXP in the MD4 algorithm. DXP processing e.g. works well with classical music, acoustic music, film and speech.

However, be sure to judge small amounts of dynamics processing at normal volumes against more processing at less volume. Compare the alternatives at equal loudness. That is of course old wisdom for any mastering engineer, but time has come to take the concept a step further: Calibrate loudness environments so metering, monitoring and processing fit together - the aim is to produce better sound quality with more consistency over a wide range of delivery formats. With the integrated approach, the ears of the mastering engineer can play a more dominant role when judging the result, not letting an uninformed sample counter get the final say.

Mastering 6000 includes a Monitor Matrix with format control and calibrated loudness settings. The line of Dynaudio AIR speakers can also provide this important anchor with 3 calibrated reference level settings, available directly on the networked remote control provided with the speakers.

Continued reading
TC Electronic Tech Library

Top